Tek ads

Two of my three 475s have not much triggering ability left. It can usually be made to sort-of work so I haven't got around to looking at it yet. I wonder if the tunnel diodes can still be had, or if it would make sense to make a module with a fast SiGe comparator to retrofit instead.

Chris

Cool!

What they don't mention is self-healing capacity. The way the story is usually told: if a crack forms in the dielectric (due to thermal or mechanical stress), the breakdown and excess leakage causes the MnO2 to oxidize the metal (and semiconducting oxides) underneath; this is usually told as localized spot heating, though I should imagine it takes place in solid state as well. The result of the reaction is insulating Mn2O3, sealing off the spot.

Obviously, too much spot heating turns into bulk heating and thermal runaway, which is why you don't want to run a tantalum at rated voltage with large dV/dt and "unlimited" current (e.g., power supply bypass).

So it's not so much reforming as self-healing, but it sort of acts the same. Leakage current starts at a moderate value and decays over time. Also explains why tants used for very low currents (timing, signal filtering, etc.) never fail -- not only is there not enough power to ignite them, they just slowly reform after stress, and all is well.

Should be possible to figure out the mechanism; spot heating should result in noisy step voltage changes, while solid state leakage conforms more to a Johnson or 1/f(^2?) noise source; given the bandwidth of being in a capacitor of course.

The best caps of all are apparently wet slug tantalums. Lower leakage-per-C than anything else, after a week's soak time. Not cheap, of course (10uF 50V, ~$70).

Tim

Interesting post from (the crowd gasps) EDN.

Cheers

Phil Hobbs

That "fix" is probably not reliable. Tantalums will fail in the field even after surviving factory test. And it's erratic between different batches of caps.

The best fix is to never put MnO2 tants across supply rails; use ceramics, alums, or polymer alums. If you really want tants, derate 3:1 on voltage.

Tantalums are very reliable if not connected to supplies that have a lot of current capability. I have incinerated tantalums across the output of LM1117 regulators.

"John Larkin" wrote in message news: snipped-for-privacy@4ax.com...

What's really handy about tants is the stable ESR and CV product. That lets you design the power bus so teensy bypasses are well damped rather than ringing between various structures.

Electrolytics have plenty of ESR, but it's unstable, and not recommended for important things like switchers and LDOs (depending on whose "Any Cap" marketing you believe in!).

Ceramics are great, but their CV is crap, especially on the larger values. Last I looked, an 0805 size X7R rated 50V still has 40% of its nominal value at rated voltage, but 1uF (everything else the same) drops to a pitiful 15%! (That was TDK's data. I don't think they really want you finding that, and the corresponding data doesn't even *exist* for their ferrite beads.) On a 3.3V rail, a pile of say 47uF 10V 1206s isn't bad, but you'll waste board space trying the same trick on, say, 24V rails.

Polymers are actually too good. With ESR as low as ceramic, but CV up there with tant, PS ripple doesn't stand a chance. But therein lies the problem: with such low impedance shorting the power supply rail, all traces to and from the caps become inductors, and the entire ground plane is a high-Q capacitor to resonate with. Sure, you could bring the damping back by adding a series resistor, but... then why poly at all?

Tim

Interesting (though I haven't used tantalums for 20 years, or so). In the 70s we had problems with tants exploding. That wasn't a big problem because with a 1000A supply behind them, they quickly disappeared. However, the interplane short that was left after the torch out the bottom just grew larger. It was a severe issue with the